How many pricing tiers do SaaS pricing pages actually show — is 3 really the norm?
We can't publish a clean tier-count distribution: of the 183 labeled desktop SaaS pricing pages Lazyweb tracks, 0 join to the structural vision data, so tier counts aren't computable as a census [1]. What the 301 detected pricing-area experiments do show is that the interesting movement is *between* 3 and 5 tiers, not a fixed 3-tier standard — real before/after diffs range from collapsing to a single hero price up to a five-step $0-to-Custom staircase [2]. Treat '3 tiers' as a common starting point companies actively test away from, not a rule.
Across 301 detected pricing-area experiments (July 2026), tier changes span from a single hero price to a five-tier $0–Custom staircase — no fixed '3-tier norm' held.
Why we won't quote a tier-count percentage
Honesty first: the 183 desktop pricing pages in our web corpus store only {screen_type, in_product} in their raw label, and none of them join to the vision-JSON table that would encode tier count [1]. Attempts to parse tier counts from experiment summaries yielded only single-digit hits (two-tier 3, three-tier 3, four-tier 1, five-tier 1) — far too sparse to publish a distribution [2]. So anyone quoting 'X% of SaaS pages use exactly 3 tiers' is guessing. The defensible answer is directional, drawn from what companies actually *changed*.
What the detected experiments show tier counts doing
Across the 301 pricing-area experiments, the observed tier-count moves go in both directions [2]:
| Observed change | Direction | Inferred rationale |
|---|---|---|
| 3-tier comparison → single '$8/mo' hero price | Fewer | Remove the plan-evaluation step for a mature audience |
| 3 paid tiers → 5-step staircase ($0, $5, $15, $30, Custom) | More | Capture hobbyists at $0–$5 instead of bouncing them at a $10 floor |
| 4 tiers → 3 (mid 'Business' tier deleted) | Fewer | Spare individuals a three-way comparison, push agencies to sales |
| 3 per-person tiers → 4 plans split into individual + team tracks | Restructure | Stop solo users parsing '/person' pricing |
These are detected UI diffs with model-inferred rationale, not measured A/B lift [2]. The pattern: teams add tiers to widen entry (free/cheap floor) and cut tiers to reduce comparison load — 3 is a midpoint, not a destination.
How to apply this
Don't anchor on '3 because everyone does 3.' Pick tier count from your funnel goal: add a $0–$5 floor tier if you're bouncing hobbyists (Framer-style five-step staircase); delete a mid tier if a three-way comparison is stalling self-serve buyers (the Creator/Pro/Business/Enterprise → 3-tier move); split into audience tracks if solo users are confused by per-seat pricing meant for teams (the Notion-style individual/team split) [2]. Each of these was a real detected change scored impact 5/5 by our model — high relative priority, though not confirmed lift.
The numbers
| Stat | Computed from |
|---|---|
| 183 | sites_screen_labels WHERE screen_type='pricing' — 183 pages across 147 companies |
| 0 of 183 | web pricing pages joining the vision-JSON structural table |
| 301 | _experiment_annotations WHERE area IN ('PRICING','PRICING TABLE') — 254 mobile + 47 web |
| two-tier 3, three-tier 3, four-tier 1, five-tier 1 | keyword parse of tier counts over experiment summaries — too sparse to publish a distribution |
Sources & citations
- [1] Lazyweb Research analysis of 183 pricing pages (desktop SaaS web corpus, 147 companies), July 2026. Labeled screen_type='pricing'; raw labels store only {screen_type,in_product} and 0/183 join the vision-JSON structural table, so tier-count distribution is not directly computable. ↩
- [2] Lazyweb Research analysis of 301 pricing-area experiments (detected before/after UI diffs), July 2026. 254 mobile PRICING + 47 web PRICING TABLE annotations; tier changes described in before/after summaries with LLM-inferred rationale, never measured A/B lift. Tier-count keyword parse yielded only single-digit hits. ↩
Source: Lazyweb Research — proprietary analysis of real, in-market app screens. Cite as Lazyweb Research, 2026-07-07.